Table 2. Summary of each included studies.
| Study | Methodology | Datasets | Number of images | Classes | Other diseases | Sensitivity | Specificity | Limitations |
|---|---|---|---|---|---|---|---|---|
| Alqudah 2019 [40] | 15-layer CNN | Duke, Mendeley, Self-built | 136187 OCT images | 4 | CNV, DME, Normal | 100% | 100% | |
| Bhatia 2019 [41] | VGG-16 | Duke, Mendeley, Noor, Self-built | 5588 OCT images | 4 | CNV, DME, Normal | 94% | 90% | 1) Ignored bad quality pictures. |
| Celebi 2022 [42] | CapsNet | Kaggle dataset, | 726 OCT images | 2 | Normal | 100% | 99% | 1)Did not study other retinal diseases; 2)Ignored bad quality pictures and patients who had other retinal diseases. |
| Dong 2022 [43] | A joint CNN detector using Yolov3 | Multicenter Self-built | 208758 FP images | 11 | DR, Glaucoma, Pathological myopia, Retinal vein occlusion, Macula hole, Epiretinal macular membrane, Hypertensive retinopathy, Myelinated fibers, Retinitis pigmentosa, Normal | 88% | 98% | 1)Only small number of retinitis pigmentosa. |
| Gour 2020 [44] | VGG-16 | ODIR | 331 FP images | 8 | Cataract, Diabetes, Glaucoma, Hyperattention, Myopia, other abnormalities, Normal | 6% | 94% | 1)The dataset contained 8 types of diseases, but with a small dataset. |
| He 2022 [45] | ResNet-50 | Duke, Mendeley | 795 OCT images | 3 | DME, Normal | 96% | 99% | 1)Only contained one other diseases. |
| Kadry 2021 [46] | VGG-16 | iChallenge-AMD database, OCTID | 3200 FP and 3200 OCT images | 2 | Non-AMD | 88% | 85% | 1)The definition of non-AMD is not clear. |
| VGG-19 | 84% | 87% | ||||||
| AlexNet, | 88% | 85% | ||||||
| ResNet-50 | 88% | 84% | ||||||
| Lee 2017 [47] | VGG-16 | Self-built | 101002 OCT images | 2 | Normal | 90% | 91% | 1)Included only images from patients who met the study criteria, and the neural network was only trained on these images; 2) This model was trained using images from a single academic center, and the external generalizability is unknown |
| Ma 2022 [48] | ResNet-34 | Self-built | 73 OCT images | 2 | Polypoidal choroidal vasculopathy | 92% | 90% | 1) Small dataset |
| Mathews 2022 [49] | A 11-layer lightweight CNN | Duke, Mendeley | 10907 OCT images | 3 | DME, Normal | 100% | 100% | 1) This study used drusen macular degeneration for AMD diagnosis; 2)Only contain one other diseases. |
| Matsuba 2019 [50] | A 7-layer CNN | Self-built | 364 OPTOS images | 2 | Normal | 100% | 97% | 1) It is difficult to acquire precise images using OPTOS when the transmission of light into the eye is impaired by an intermediate translucent zone; 2) Most AMD patients accept treatment which may cause diagnostic error 3) Did not study other retinal diseases. |
| Motozawa 2019 [51] | An 18-layer CNN | Self-built | 169 OCT images | 2 | Normal | 99% | 100% | 1) Excluded low quality images and patients who had other concomitant diseases; 2) Did not study other retinal diseases. |
| Takhchidi 2021 [52] | ResNet-50 | Self-built | 1200 FP images | 2 | Normal | 90% | 86% | 1) Did not study other retinal diseases. |
| Tan 2018 [53] | A 14-layer CNN | Self-built | 1110 FP images | 2 | Normal | 96% | 94% | 1) Did not study other retinal diseases. |
| Thomas 2021 [54] | A 19-layer CNN | Mendeley, Duke, Noor, OCTID | 1139 OCT images | 2 | Normal | 99% | 100% | 1) Did not study other retinal diseases. |
| Wang 2019 [55] | DenseNet-121 | Duke, Noor | 8315 OCT images | 3 | DME, Normal | 96% in Duke, 95% in Noor | 95% in Duke, 95% in Noor | 1) Only contained one other diseases. |
| ResNet-50 | 97% in Duke, 100% in Noor | 100% in Duke, 99% in Noor | ||||||
| ResNext-101 | 100% in Duke, 99% in Noor | 100% in Duke, 95% in Noor | ||||||
| DPN-92 | 97% in Duke, 100% in Noor | 97% in Duke, 99% in Noor | ||||||
| CliqueNet-10 | 99% in Duke, 93% in Noor | 99% in Duke, 98% in Noor | ||||||
| Yoo 2018 [56] | VGG-19 | Project Macula | 83 FP and 83 OCT images | 2 | Normal | 84% | 59% | 1) Did not study other retinal diseases; 2) Small datasets; |
| Zapata 2020 [57] | A 24-layer CNN | Optretina’s tagged dataset | 306302 FP images and OCT images | 2 | Glaucomatous optic neuropathy | 83% | 89% | 1.No clear number of OCT or FP images. |